Running Mercurial test suite on Windows

Erik Zielke ez at
Thu Dec 9 02:52:15 CST 2010

Den 08-12-2010 23:40, Patrick Mézard skrev:
> Le 08/12/10 17:42, Erik Zielke a écrit :
>> Hi
>> I have tried to run the test suite on Windows, according to the instructions at
>> I updated my Mercurial clone back to where the queue was last applied, and applied the windows testing queue there.
>> I use Python 2.6, Windows 7 and run the tests from the command prompt. I wasn't sure whether the tests should be run from cygwin, but then there was a problem with paths. Someone else mentioned something about the same thing (I think) on the mailing list a few days ago. So I guess that one should use cmd to run the tests.
> No but you need cygwin configured with LF output on your PATH.
Yes, the guide also says that, but can't find info on configuring cygwin 
to that. How is that done?

>> But I ran into some issues with, e.g.:
>> - confusion over the two different paths separators
>> - exit codes: a lot of places tests report 127 instead of 255 as exit code.
>> Can these problems be because I have set up the test suite incorrectly, because they occur in a lot of the tests for me?
> First, I have not spent a lot of time in the past six months updating the patch queue, so you can expect a lot of failures. I have fixed a couple of them recently but I lack of time to do that thouroughly. Secondly, the test suite never passed completely under Windows.

Yes. I have seen there are some patches in the queue with tofix guards. 
It did also understand that the test suite didn't pass completely under 
Windows, so thought I would try to see if it would be possible for me to 
make more of the tests run correctly under Windows.

> When possible, I deal with paths separators by automatically rewriting the output of some shell commands and several mercurial commands directly in pysh. I do that when there is either too much breakage compared to the loss of coverage or when the path separator nature is unimportant.

> Something I should fix is I still run the tests nightly but they never finish cleanly because one of the tests causes to fail when parsing the output (it might be caused by one of my own patch). If I correct this you will have some kind of test result reference here:


> --
> Patrick Mézard

Best regards,
Erik Zielke

More information about the Mercurial-devel mailing list